Virulence factor expression patterns in Pseudomonas aeruginosa strains from infants with

Transcription

1 Virulence factor expression patterns in Pseudomonas aeruginosa strains from infants with cystic fibrosis Jim Manos 1 *, Honghua Hu 1 *, Barbara R. Rose 1, Claire E. Wainwright, Iryna B. Zablotska, Joyce Cheney, Lynne Turnbull, Cynthia B. Whitchurch, Keith Grimwood, Christopher Harmer 1, Snehal N. Anuj, Colin Harbour 1 and the ACFBAL study group. 1 1 Sydney CF Research Group, Department of Infectious Diseases and Immunology, University of Sydney, Sydney, NSW, 00, Australia Queensland Children s Medical Research Institute, Royal Children s Hospital, The University of Queensland, Brisbane, Australia The Kirby Institute, University of New South Wales, Darlinghurst NSW, 0, Australia The ithree Institute, University of Technology Sydney, Sydney, NSW, 00, Australia * Authors contributed equally Corresponding Author. Address: Department of Infectious Diseases and Immunology, Blackburn Building, University of Sydney, NSW 00 Australia. Ph: Fax: jim.manos@sydney.edu.au Current address: Australian School of Advanced Medicine, Macquarie University, NSW,, Australia 1 Running Title: Phenotype of P. aeruginosa from infants with cystic fibrosis Key words: Pseudomonas aeruginosa, infection, phenotype, virulence factor, cystic fibrosis, infant/child 1

2 ABSTRACT Pseudomonas aeruginosa is the leading cause of morbidity and mortality in cystic fibrosis (CF). This study examines the role of organism-specific factors in the pathogenesis of very early P. aeruginosa infection in the CF airway. One-hundred-and-sixty-eight longitudinally collected P. aeruginosa isolates from children diagnosed with CF following newborn screening were genotyped by pulsed-field gel electrophoresis and phenotyped for 1 virulence factors. Associations between virulence factors and gender, exacerbation, persistence, timing of infection and infection site were assessed using multivariate regression analysis. Ninety-two strains were identified. Persistent strains showed significantly lower levels for pyoverdine, rhamnolipid, haemolysin, total protease, and swimming and twitching motility, than strains eradicated by aggressive antibiotic treatments. Initial strains had higher levels of virulence factors, and significantly higher phospholipase C, than subsequent genotypically-different strains at their initial isolation. Strains from males had significantly lower pyoverdine and swimming motility than females. Colony size was significantly smaller in strains isolated during exacerbation than those isolated during non-exacerbation periods. All virulence factors were higher and swimming motility significantly higher, in strains from bronchoalveolar-lavage (BAL) and oropharyngeal sites than BAL-alone. Using unadjusted regression modelling, age at initial infection and age at strain isolation showed U-shaped profiles for most virulence factors. Among subsequent strains, the longer the time since initial infection, the lower the level of most virulence factors. This study provides new insight into virulence factors underpinning impaired airway clearance seen in CF infants despite aggressive antibiotic therapy. This information will be important in the development of new strategies to reduce the impact of P. aeruginosa in CF.

3 INTRODUCTION Pseudomonas aeruginosa airway infection of young children with cystic fibrosis (CF) is typically intermittent, involving multiple non-mucoid strains that are sensitive to anti-pseudomonal antibiotics. However, one strain eventually becomes established and by adolescence almost 0% of CF patients are chronically-infected, resulting in increased morbidity and mortality [1] Many host factors affect the course of infection in the CF lung [], but there is limited understanding about organism-specific traits affecting infectivity, severity of infection and persistence. The environment is the main source for acquiring P. aeruginosa, suggesting that environmental isolates have all the virulence determinants to initiate infection and cause disease []. P. aeruginosa adapts to long-term survival in the CF lung by overproducing alginate, forming biofilms and down-regulating virulence factors [-]. Tracking of adaptive mutations in isolates from chronically-infected adults showed defects in swimming and twitching motility, pyocyanin secretion and biofilm formation []. However, P. aeruginosa strains infecting infants and young children have not been systematically examined to establish the characteristics underpinning pathogenesis of early infection in the CF lung Here we present the virulence factor phenotype of 1 P. aeruginosa isolates collected longitudinally over the first five-years of life from infants diagnosed with CF after newborn screening and classified by genotyping as distinct strains. The Type Secretion System profiles of this cohort were published previously []. Relationships were sought between virulence factor expression and clinical factors.

4 MATERIALS AND METHODS Clinical isolates Investigations were conducted on 1 P. aeruginosa isolates from children aged - (mean.) months collected within the Australasian Cystic Fibrosis Bronchoalveolar Lavage (ACFBAL) randomised controlled trial []. Isolates came from specimens: bronchoalveolar lavage (BAL) fluid ( children); oropharyngeal (OP) swabs ( children) and sputa ( children post-acfbal study). On average,. (range 1-1) isolates were tested per child. Ethics Committees of all participating hospitals approved the study The ACFBAL study randomised infants diagnosed with CF after newborn screening to either BALdirected therapy or standard care using clinical judgment and OP cultures []. BAL-directed infants underwent BAL at enrolment before six-months of age, and then with exacerbations requiring hospitalisation, if OP cultures grew P. aeruginosa, and after P. aeruginosa eradication therapy. All subjects had OP cultures taken during exacerbations and after eradication therapy, while each underwent BAL at age five. Children with positive P. aeruginosa BAL ( colony-forming units/ml) or OP (standard group only) cultures received two-weeks of intravenous tobramycin with either ticarcillin-clavulanate, or ceftazidime, followed by four-weeks of oral ciprofloxacin and eight weeks of nebulised tobramycin inhalation solution []. After treatment, BAL-directed children had further BAL and OP cultures, while those receiving standard care had OP cultures alone. If P. aeruginosa persisted, the eradication protocol was repeated, but if this also failed the child was categorised as chronically-infected. Samples were processed by hospital laboratories where P. aeruginosa identification was undertaken using standard techniques and interpretative criteria [].

5 Genotypic testing Isolates were genotyped using pulsed-field gel electrophoresis (PFGE) following SpeI digestion []. Analyses were performed using cluster analysis software (GelComparII TM, Applied Maths, Belgium) and the criteria of Tenover et al []. Phenotypic tests i. Pyocyanin Overnight P. aeruginosa grown in Cation-Adjusted Mueller-Hinton Broth (CAMHB) (Oxoid Ltd, Australia) were scored (0-) for intensity of green pigment: 0 = no pigment, 1 = light green, = moderate green, = dark green, = very dark green [1]. ii. Pyoverdine µl of overnight CAMHB cultures were spotted onto Kings B agar (Oxoid), 1 1 dried and incubated at o C. Plates were viewed under a transilluminator after h and scored on the intensity of the fluorescent yellow pigment produced by pyoverdine using a graded system: 0= no fluorescence, 1 = lightly fluorescent yellow, = moderately fluorescent yellow, = strongly fluorescent yellow, = very strongly fluorescent yellow [1]. 1 iii. Haemolysin µl of overnight CAMHB culture was spotted onto Columbia Horse Blood Agar (Oxoid), dried and incubated for 0h at o C. Two perpendicular diameters of clearance zones around colonies were measured and the area (πr ) calculated (mm ). This method was utilised in iv to viii below. 1 iv. Total Protease µl of overnight CAMHB culture were spotted onto a 1%v/v skim milk agar plate dried and incubated for h at o C. v. Elastase µl of overnight CAMHB culture were spotted onto ml elastin agar plates, dried and incubated for h at o C. vi. Phospholipase C µl of overnight CAMHB culture were spotted onto egg yolk agar plates, dried and incubated for 0h at o C [1-1]. vii. Rhamnolipid µl of overnight CAMHB culture were spotted onto M-based agar plates, supplemented with 0.% (v/v) glucose, mm MgSO, 0.0% (v/v) tyrosine, 0.000% (v/v)

6 methylene blue, 0.0% (v/v) and cetyltrimethylammonium bromide dried and incubated for h at o C. Colonies were scored as positive by the presence of a violet halo around the colony. viii. Twitching and swimming motilities Bacterial patches from 1.%w/v Luria plates were scraped with a toothpick and stab inoculated into ml (1%w/v Luria-broth) agar. Twitching motility (light haze of growth at the agar/plate interface) was measured after h at o C. For swimming, toothpick scrapings were spotted onto centre of a swim (0.%w/v Luria-broth) plate and swim diameters measured after overnight at 0 o C. ix. Swarming motility Pre-warmed swarm plates (0.%w/v nutrient broth, 0.%w/v agar and %w/v glucose) were inoculated with a colony tooth-picked from a swim plate (above). After h at o C, swarming motility was identified as spreading growth from the inoculation site and graded 0- where; 0 = no motility, 1= very limited motility. = fair motility, = strongly motile, = very strongly motile. x. Biofilm mass 1 0 cells were inoculated into wells of a microtitre plate containing 0µl CAMHB and plates incubated for h at o C. Biofilm mass was stained with 1%v/v crystal violet, ethanol-extracted and absorbance read at O D. Readings were then normalised as a percentage of P. aeruginosa PAO1 biofilm grown concurrently. 1 xii. Colony size 0µl aliquots of a - dilution of overnight P. aeruginosa culture were spread 1 onto CAMHB plates and incubated at o C. Colony diameter was measured in mm after h. 1 xiii. Mucoidy Cells were scraped from 1.%w/v Luria-broth patch plate, then patched onto 0 %v/v glycerol MacConkey plates (Oxoid) and scored (+/-) for mucoidy after h at o C. 1 Statistical analysis Patients were followed longitudinally. Each isolate was tested ( ) for each outcome of interest. Nine continuous and three categorical measures of virulence (measured using semi-quantitative scores from 0-to-) were analysed. These scores were categorised into binary variables to assess the

7 likelihood of having a score of or versus any score below. As only four isolates were mucoid, statistical analysis was not conducted on the fourth categorical measure (mucoidy). The associations between a set of independent factors and each of the virulence measures were assessed using multivariate regression analysis. The independent factors were gender, exacerbation at sample collection time, initial vs subsequent strain infection, persistent strain (defined as one detected continuously or intermittently for at least three-months) and site of sample isolation (BAL only, BAL+OP, OP-only). 1 1 We further investigated whether: i) the child s age at acquisition of first infection, ii) the age of the child at the collection of a strain and iii) time since acquisition of the first strain, had any effect on the outcomes of interest. The associations between these three time-related variables and each of the virulence measures were estimated separately using unadjusted regression analysis Population-averaged panel-data regression models and generalized estimation equations (GEE) were used to assess the associations of interest. Linear or logistic regression was used depending on the measure of the outcome of interest. Regression analyses were executed using STATA 1.0 (StataCorp, College Station, TX, USA).

8 RESULTS The children fell into three categories. Twenty-eight had only one strain detected on a single occasion; had one strain detected on multiple occasions and the remaining 0 had multiple strains detected on multiple occasions. Genotype Ninety-two strains were identified by PFGE. Of these, 0 strains had unique pulsotypes, while the remaining two strains had macrorestriction patterns indistinguishable from Aust-01 and Aust-0, which are highly-prevalent strains found frequently in older, chronically infected patients attending Australian CF clinics [1]. Only one patient had Aust-01, and another two had the Aust-0 strain identified. Since transmissibility and other properties of these clonal strains may have biased results, all analyses were repeated without them and results remained comparable (data not shown) Of the strains, (.%) were initial infecting strains detected in children at an average age of. (range -, standard deviation 1.) months. Seventy-three strains (.%) came from an exacerbation, strains (.%) were persistent (1 were persistent from the initial infection and eight from a subsequent infection), (.1%) cleared, while the persistence of (1.0%) strains was unknown as genotyping was limited to the available sample set. Thirty-five strains were from the upper airways (OP) only, 1 from the lower airways collected by BAL and 1 were found in both sites (eight strains from follow-up sputa were excluded from analyses for site, as the quality of these specimens was unknown and they may have contained both upper and lower airway isolates). Phenotype Pyocyanin was scored at or in 0.% of the tests, pyoverdine was scored at or in 1.% of the tests and swarming was scored at or in 1.% of the tests. The distributions of all continuous measures of interest are presented in Table 1.

9 Persistent strains versus cleared strains Multivariate analysis (Table ) showed that the levels of six virulence factors: pyoverdine, rhamnolipid, haemolysin, total protease and swimming and twitching motilities were significantly lower in persistent strains compared to cleared strains, while another six factors; pyocyanin, elastase, phospholipase C, biofilm mass, colony size and swarming motility showed non-significant reductions. No factor was significantly higher in persistent strains compared to cleared strains Effects of gender, timing of isolation, site of infection and exacerbation on phenotype With respect to gender, pyoverdine expression and swimming motility were significantly lower in strains from males compared to females (Table ). In comparisons of initial and subsequent (genotypically-different) strains at their first isolation, initial strains had higher expression of all virulence factors, except pyocyanin, and this reached statistical significance for phospholipase C. All virulence factors showed an increase in strains from both OP and BAL compared to BAL-only, while swimming motility was significantly higher. Only smaller colony size was associated with exacerbation Influence of age of child and time of isolation on phenotype Table shows changes in virulence factor expression and Figure 1 shows the changes for the eight clearance-zone based measurements, with respect to: age at initial strain isolation, age at isolation of a strain and time since initial strain isolation. 1 Age at initial strain isolation: Phospholipase C, haemolysin and total protease levels were significantly lower in initial strains acquired at age 1- months compared with 1-months, but were not different at age >-months compared with 1-months (Table, part 1). Pyoverdine and swarming motility were lower in initial strains acquired at - months compared with 1- months, but were the same at months as at 1 months. Biofilm mass increased with age of

10 initial strain isolation, becoming significant at - months. Of the eight virulence factors measured by area of clearance zone, all except biofilm mass produced a non-linear, non-monotonic U-shaped curve (Figure 1A). Age at isolation of a strain: Expression of pyoverdine and haemolysin, and colony size were significantly lower in strains from children aged --months compared to children aged 1months. Total protease expression was significantly lower in both 1- and --month-old children compared with children aged 1months (Table, part ). Of the eight virulence factors measured by area of clearance zone (Figure 1B) all except swimming and twitching motility gave a U-shaped effect Time since isolation of initial strain; The longer the time, the lower the level of most virulence factors (Table, part ). Significant decreases occurred for biofilm mass and swimming motility by age -months, for twitching motility and colony size by -months and for elastase, phospholipase C, haemolysin and total protease by >-months. Figure 1C, shows an initial fall, followed by a stable period, with a sharp decline at >-months, for all factors except rhamnolipid. The presence of subsequent infections does not explain the sharp decline in virulence observed in the month s group. Subsequent strains were isolated from patients between and months after initial strains (mean=. months, median=1 months), meaning that most subsequent infections occurred before months. 1

11 DISCUSSION P. aeruginosa virulence factors are usually associated with acute infection within the host. The complex relationships between virulence factor expression and pathogenesis of airway infection in CF have been investigated mainly in chronically infected adults [1-1]. Our study is the first to comprehensively examine P. aeruginosa virulence factor expression profiles in CF infants and young children. Virulence versus persistence, initial/subsequent strain, gender, exacerbation and site of infection This study provides the first evidence that virulence factor expression profiles of P aeruginosa strains at their first isolation in the CF airway of young children may help predict the ability of the strain to persist in the face of the immune response and antibiotic therapy. Expressions of pyoverdine, rhamnolipid, haemolysin, total protease and swimming and twitching motilities were all significantly lower in persister than clearer strains and this finding was independent of potential confounders such as site of isolation (upper or lower airway) and whether the airway was P. aeruginosa naive or potentially damaged by a previous P. aeruginosa strain. Since these persister strains were not clonally related to each other or to frequent clones circulating in Australian clinics they are assumed to have been environmentally acquired. P. aeruginosa is known to down-regulate these virulence factors as a survival mechanism in the CF lung [1, 0]. However it seems unlikely that such adaptations begin so early in the infection process. Perhaps instead some environmental strains first colonise the sinuses where adaptation to the local microenvironment takes place before aspiration into the lower airways occurs. Studies show pyoverdine-negative mutants increase with lung colonisation time [1], which would indicate that persisters do not have a greater requirement for pyoverdine. Rhamnolipid is a biosurfactant involved in protecting cells against oxidative stress [] and in increased swarming []. It is also cytotoxic to eukaryotic cells, decreases liquid surface tension and facilitates access to nutrients within biofilms [-]. Since these functions are important in long-term bacterial survival, the down regulation of rhamnolipid is noteworthy. Decreased haemolysin expression has been correlated with increased mucoidy [], however all

12 persistent strains in our study were non-mucoid. Some bacterial proteases promote mucoidy [], so strains may begin to display mucoidy as persistence continues. The overall virulence profiles of initial strains were higher than those of subsequent strains at their first isolation. In particular, phospholipase C was significantly higher in initial strains. Phospholipase C is a heat-labile lecithin-degrading haemolysin. The phospholipase C/sphingomyelinase gene-pair acts as an inhibitor of lung pulmonary surfactant [], thus phospholipase C may be required to establish the initial, but not subsequent infection Associations with gender proved interesting. Previous reports had indicated high virulence factor expression during early CF lung disease played a role in the worse prognosis of females [], however there was no evidence of this in our analysis. The significance of the lower pyoverdine expression and lower swimming motility in strains from males is unknown. There was a lack of association between virulence factor expression and exacerbation other than for colony size, and this was at odds with our early study of CF where higher extracellular enzyme levels were found in hospitalised than non-hospitalised CF adolescent patients with chronic P. aeruginosa infection [0]. We found no significant associations between expression and site of infection, except for swimming motility being greater in strains from both BAL and OP than those from BAL-alone. These conflicting results could reflect differences in age, disease stage and the small number of nonexacerbation samples. 1 Flagella and pili-driven twitching motility are required for biofilm formation [1]. Alternatively, swarming motility is inversely proportional to biofilm formation []. Our finding that swarming motility and biofilm mass were not significantly different, while twitching and swimming motility were significantly reduced in persisters highlights differences between early persisters in young children and chronic infection in CF adults where biofilm formation is a distinctive feature []. 1

13 1 1 1 Influence of age of child and time of isolation on phenotype Particularly noteworthy were associations between age and virulence factor expression (Table and Figures 1A 1B 1C). The initial reduction in virulence factor production with increasing age of initial infection (Figure 1A) could be an adaptation to the host immune response. However, the increase in almost all factors in initial strains in the older age groups was unexpected. This was particularly marked for haemolysin, which appears independent of age of acquiring the initial strain. A U- shaped curve was also evident in the analyses of virulence by age at isolation of a strain (Figure 1B), with strains from children >-months showing a modest rise after an initial fall for all factors except swimming and twitching motilities, and haemolysin again showing a significant fall at - months followed by a rise at months. While we are unsure of the reasons for this U-shaped effect, the fall in virulence factor expression seen in new P. aeruginosa isolates from older children, could offer a partial explanation. As the sinuses develop and grow in CF children, they may act as a reservoir for initial and newly acquired P. aeruginosa strains, where adaptation to the lung environment may already begin prior to them infecting lower airways [-] In contrast, results of the analyses examining time since isolation of initial strain (Figure 1C), which showed significant decreased expression for eight of the twelve virulence factors by the month time period. The significant decreases start from the earliest time point (1- months) in the case of biofilm mass and swimming. These results are comparable to those found by a study of P. aeruginosa isolates from CF children of undisclosed ages [], except that this trend was evident at months compared with over >-months in our study. Interestingly, four virulence factors, (pyocyanin, pyoverdine, swarming and rhamnolipid) remained virtually unchanged over the entire period. This study provides important new information on factors influencing virulence factor expression of P aeruginosa infecting the CF airway in the first years of life. While it is generally accepted that P. 1

14 aeruginosa loses its virulence in the transition to chronic infection in CF, our results which show a a diverse range of up and downregulation as well as maintenance of expression amongst the twelve virulence factors over time, challenge this view. It seems likely that there is considerable diversity within microbial population evolution, with some strains retaining or enhancing virulence and others losing virulence over time. Our data suggest that this process begins within or so months after initial infection. Our findings may help in understanding which virulence factors are important in establishing infection in CF infants where, despite aggressive antibiotic therapy, impaired airway clearance mechanisms exist. This is turn will be important in developing new more effective strategies to reduce the impact of P. aeruginosa on morbidity and mortality in CF. 1

15 ACKNOWLEDGEMENTS This work was supported by grants from the NHMRC (# ) and the Australian Cystic Fibrosis Trust. Cynthia B. Whitchurch was supported by a NHMRC Career Development Award and a NHMRC Senior Research Fellowship. 1

16 CONFLICT OF INTEREST All authors declare that they have no conflict of interest. 1

17 TABLES AND FIGURES Table 1: Statistical distributions of all continuous measures of interest Variable Mean Std. Dev. Min Max Elastase (mm ) Rhamnolipid (mm ) Haemolysin (mm ) Total Protease (mm ) Phospholipase C (mm ) Biofilm mass (% of PAO1 biomass)..0.. Swimming (mm ) Twitching (mm ) Colony Size (diameter in mm)

18 Table : Multivariate analysis of associations between a set of independent variables and individual virulence factors Independent variables Pyoverdine Pyocyanin Swarming Elastase OR a (% CI) OR a (% CI) OR a (% CI) AEC b (% CI) Gender: Male vs Female 0. (0.-0.)* 0. (0.-.) 0.0 ( ) -. (-1.-.) Exacerbation: Yes vs No 0. (0.0-1.) 1. (0.-.1) 0. (0.1-.0) -. (-.-.) Strain: Initial vs Subsequent 1. (0.-.) 0. (0.1-1.) 1. (0.-.) 1. (-.-.) Strain: Persister vs clearer 0.1 (0.0-0.)** 0. (0.-.1) 0.0 (0.1-1.) -.0 (-.0-.) Sample: OP vs BAL 1. (0.-.1) 1. (0.-.) 0.0 (0.0-.0). (-1.-.) Sample: (BAL+OP) vs BAL. (0.-1.). (0.0-1.) 1. (0.1-.0).0 (-1.-.0) Independent variables Total Protease Rhamnolipid PLC c Haemolysin AEC b (% CI) AEC b (% CI) AEC b (% CI) AEC b (% CI) Gender: Male vs Female. (-.-1.) -. (-1.0-.).0 ( ) -. (-.-1.) Exacerbation: Yes vs No -. (-1.1-.) -1. (-1.-.) -.0 (-.1-.1) -1.1 (-.-1.) Strain: Initial vs Subsequent. (-.0-.). (-.1-.).0 (.-.)* 1. (-1.1-.) Strain: Persister vs clearer -1. ( )* -1. (-.--0.)* -. (-.-.) -. (-.--.0)* Sample: OP vs BAL.0 ( ) -. (-1.-.) -. (.1-1.0) -.1 (-.-.0) Sample: (BAL+OP) vs BAL. (-.-.). (-.1-.1). (-0.-.) 1. (-1.-.) Independent variables Swimming Twitching Biofilm Mass Colony Size AEC b (% CI) AEC b (% CI) AEC b (% CI) AEC b (% CI) Gender: Male vs Female -1. (-.--.)* -.1 ( ) 1.1 (-1.-1.) 0. ( ) Exacerbation: Yes vs No 0. (-1.-1.). ( ) -1. (-.-.1) -0. ( )* Strain: Initial vs Subsequent 1.0 (-1.0-.) 1.1 (-0.-.) 1. (-.-1.) 0. (-0.-0.) Strain: Persister vs clearer -. (-.1--.)*** -. (-.--.)*** -1. (-.1-.) -0.1 ( )

19 Table : Change in virulence factor expression over time Part 1: Age at initial strain isolation Virulence factor Units <1mths 1-mths pvalue a -mths pvalue a +mths pvalue a Pyocyanin % Pyoverdine % Swarming % * Elastase mm Rhamnolipid mm PLC b mm.. 0.0* Haemolysin mm *. 0.00** Total Protease mm * Biofilm Mass % c ** Swimming mm Twitching mm Colony Size mm Part : Age at isolation of a strain Virulence factor Units <1mths 1-mths pvalue a -mths pvalue a +mths pvalue a Pyocyanin % Pyoverdine % **. 0.1 Swarming % Elastase mm Rhamnolipid mm PLC b mm Haemolysin mm ** Total Protease mm * * Biofilm Mass % c Swimming mm Twitching mm * Colony Size mm *. 0.00** 1

20 Part : Time since isolation of initial strain Virulence factor Units 0mths 1-mths pvalue a 1-mths pvalue a -mths pvalue a +mths pvalue a Pyocyanin % Pyoverdine % Swarming % Elastase mm *** Rhamnolipid mm PLC b mm ** Haemolysin mm * Total Protease mm * Biofilm Mass % c * Swimming mm * Twitching mm * Colony Size mm * *** a Wald statistic; p values obtained from unadjusted regression analysis b Phospholipase C c % of P. aeruginosa PAO1 biofilm grown concurrently * p<0.0; ** p<0.01; ***p<0.001; borderline p<0.0 0